Forecasting Electoral Violence in 2024 and 2025

• In Event: Political Forecasting Methods: Challenges and Potential Solutions

Abstract

Electoral violence remains a significant challenge to democratic processes worldwide, often undermining the legitimacy and fairness of elections. This paper introduces a novel predictive model that employs machine learning techniques to forecast the likelihood of electoral violence on the global level, on a three-level ordinal scale: no, moderate, and severe electoral violence. The level of electoral violence is calculated from an average of the government and non-government use of electoral violence as coded by the Varieties of Democracy project. Our methodology harnesses a comprehensive dataset comprising various structural factors such as economic indicators, political stabilit, historical violence records, and digital vulnerability.

We employ several different machine learning algorithms, including Random Forest and Gradient Boosting machines, to capture the complex, non-linear relationships inherent in the data. We then train several different specifications of models, i.e. features and algorithms, and weigh these using a genetic algorithm to find the optimal ensemble forecast.

Our findings reveal that the model effectively discriminates between the different levels of risk, with a high degree of predictive accuracy. The results underscore the importance of specific structural factors, such as economic inequality and political repression, but especially point towards the importance of the legacy and long-lasting effects of previous rounds of electoral violence.

This research contributes to the field of political violence prediction by providing a scalable, data-driven approach applicable to all countries. It not only advances academic understanding but also offers practical tools for policymakers and international organizations to identify potential hotspots of electoral violence, facilitating timely and targeted interventions to uphold electoral integrity and democracy. In addition, the model can be used to find cases ripe for breaking the cycle of recurring electoral violence. The model's global applicability and methodological rigor make it a valuable asset in the ongoing efforts to mitigate electoral violence worldwide.

Authors

• David Randahl, Department of Peace and Conflict Research, Uppsala University

• Maxine Leis

• Tim Gåsste, Uppsala University

• Staffan I. Lindberg, V-Dem Institute, University of Gothenburg

• Håvard Hegre, Peace Research Institute Oslo

• Hanne Fjelde, Uppsala University

• Steven Lloyd Wilson, University of Nevada, Reno